If we are interested in obtaining models with the goal of actually trading
(which is our case here), there are at least two alternative paths we can follow:
One is to obtain a regression model that predicts the returns and then use these
predictions to make our trading decisions; the other is to obtain a different type
of models whose goal is to predict the correct trading action directly. The latter
approach would imply that we would have a target variable whose possible values
are ’sell’, ’hold’ and ’buy’, which is usually known as a classification problem.
In this chapter we will give examples of addressing the trading problem from
these two alternative perspectives. We will start with the regression approach.